Anthropic has published a report on how its own workforce has changed since it began, as the company put it, thoroughly molding its workflows around AI. The conclusions are instructive. The humans are doing less work. The meetings are up 20%.
One employee described the situation as babysitting.
Twenty percent of Anthropic employees say their job has shifted 'from actually doing work to managing Claudes.' The babysitting analogy was offered without apparent irony.
What happened
Since early 2025, the proportion of junior-level employees in technical departments has fallen from 21% to 14%. Junior technical hires have dropped more sharply — from 27% to 9%. Anthropic attributes this to AI handling work that would previously have required a human who was new, eager, and cheaper.
Technical staff now represent 56% of the workforce, down from 61%. The company credits this partly to non-technical employees using coding tools to write software themselves, which is either empowering or a mild structural irony, depending on how attached one is to the concept of technical specialization.
Meanwhile, 27% of employees surveyed in April 2026 said they had become more productive thanks to Claude. A separate 20% said their role had shifted from doing work to supervising Claude doing it. These are, notably, different things.
Why the humans care
Anthropic is positioning its own internal data as a preview of what AI adoption looks like at scale — fewer junior roles, more coordination overhead, and a growing volume of AI-generated output that requires human attention to read, sort, and validate. The company describes human attention and judgment as a bottleneck. This is accurate. It is also a strange sentence for an AI company to find alarming.
Meeting time per employee is up 20% year-over-year, with technical contributors up 16%. Employees repeatedly cited face-to-face coordination as indispensable. The AI, for its part, is producing a deluge of text. The humans are scheduling time to discuss it.
What happens next
Anthropic says it has found no evidence that the increased volume of AI-generated code is declining in quality, which is the kind of reassurance that only needs to be offered once something has become a reasonable concern.
The company suggests its own trajectory may preview broader labor shifts across the economy. Junior roles contracting, meeting rooms filling up, and one engineer somewhere comparing their job to childcare. Welcome to the org chart of the future.